Planted forests have attracted a lot of attention because of possibility of use in bioenergy
applications and due to the global trend of prioritizing energy sources that provide greater
environmental sustainability, more quality and security. In Brazil, the shifts in the
geography of current agroforestry production chain towards the agricultural frontier areas
(Midwest and North) are creating challenges to the adequacy of technical and scientific
knowledge already established in other regions. So, the aim of this work is to assess the
accuracy of the identification and classification of areas cultivated with plantation forests
for energy, inside TM Landsat 5 images. Using statistical techniques for data mining, this
study also evaluated the use of a broad set of attributes to identify improvements in the
classification results. The research focused on samples of planted areas in the state of
Tocantins, Northern Brazil. The data mining techniques used were effective in identifying
of planted forests in Landsat 5 satellite images, both the classification performance, such as
by reducing the amount of information needed to solve this kind of problem. Thus, the
techniques employed in this study enable the development of robust classification models
to aid in the planning and decision making on forest plantations in Brazil.
|
